We now support working with multiple buckets for output (meta)data. This allowed result data download to also work better during failover.
This however forced us to fetch result metadata from ES index, because the result metadata contains the asset locations.

Add a string field to job metadata, which points to result metadata.

I think this can be multiple things in the future:

• a legacy job_metadata.json file
• a static stac collection file
• a STAC collection exposed via API

If we do this, then we can switch back to reading result metadata from file instead of ES, and drastically reduce the size of job metadata index in elasticsearch, which would help the migration to separate job registry.